Data & Insights: Bittensor's Tense Situation; HIP3 Records

Bittensor’s Templar Conundrum

• Bittensor's TAO token fell by over ~25% last Friday, after Covenant AI, the team behind Templar (Subnet 3), announced it was leaving the network.
  • For context, Covenant AI had completed Covenant-72B last month, the largest decentralized LLM pre-training run in history.
  • That 72-billion-parameter model, trained across 70+ nodes, scored 67.1 on MMLU zero-shot, beating Meta's LLaMA-2-70B.
  • It was the single most credible proof that decentralized AI training could actually work at scale, receiving acknowledgment and praise from the likes of Jensen Huang and Chamath Palihapitiya.
  • This brought significant attention to Bittensor, leading to TAO's price doubling in March.
• However, last week, Covenant AI's founder called out the Bittensor network for its lack of decentralization
  • He cited on-chain evidence that 38 of 41 Bittensor network upgrades between 2023 and 2026 originated from infrastructure that Bittensor's co-founder controlled directly.
  • This matters because Bittensor's core value proposition, that decentralized networks can produce frontier AI without centralized control, is precisely what Covenant AI called into question, which is a big narrative blow to Bittensor’s core thesis.
  • At the same time, the Templar team sold 37,000 TAO, worth roughly $10M, onto the open market.
• The GMCI AI Index, where TAO carries a ~25% weighting, fell from its mid-March local high of ~56 down to ~51 as of Sunday.
  • Bittensor subnet tokens related to the Templar team, mainly Templar’s SN3, Basilica’s SN39 and Grail’s SN81, all fell by over 50%.
• The next step in this saga is to watch whether Bittensor's remaining subnets can produce anything close to Covenant-72B quality.
  • On the other hand, it will be worth tracking where the Covenant AI team lands next, as that destination may end up being the real beneficiary of everything Templar proved was possible.

Hyper-accelerated Growth

• Open interest (OI) on Hyperliquid's HIP-3 markets hit a record high of $2.38 billion last week.
  • As of Monday morning, OI sits at $2.1B, still near its highs despite a 12% pullback likely tied to broader risk-off moves. 
  • HIP-3’s growth trajectory has been staggering, from just ~$280M in OI at the beginning of the year to crossing $1B less than a month later and $2B by the end of the quarter.
  • Year-to-date, HIP-3 OI is up ~580%
• TradeXYZ alone accounts for over 90% of all HIP-3 OI.
  • On the Hyperliquid platform itself, only 3 of the top 10 markets by volume are crypto pairs. The rest are tokenized equity and commodity futures such as Nasdaq proxies, oil, gold, silver, and the S&P 500.
  • In case HIP-3’s value prop wasn’t clear enough, it's that traders want exposure to equities and commodities outside traditional market hours, and HIP-3 provides that without intermediaries and with crypto-native settlement.
  • Existing tradfi incumbents such as CME, ICE, and Eurex don't offer 24/7 equity futures, while Robinhood and Interactive Brokers don't settle on-chain.
  • Seeing the clear demand and product-market fit of HIP-3’s 24/7 trading, tradfi exchanges will likely respond by extending trading hours themselves.
• The $5B OI mark is where things get interesting for HIP-3.
  • At that level, HIP-3 starts generating enough volume and liquidity to attract market makers who currently operate on CME and CBOE.
  • Moreover, there's a good possibility that HIP-3 expands beyond perps into spot tokenized equities, which would represent a much more direct challenge to traditional stock exchanges and would also almost certainly accelerate regulatory response.

Hold on to your AUM

• Spot Ethereum ETF AUM has declined sharply from its peak of $28B in October 2025 to $10.9B currently, a roughly 61% drawdown driven by a combination of net outflows and ETH's price falling from approximately $4,000 to $2,200 over the same period. 
  • While there are two forces moving AUM simultaneously, it's clear that the fervor for ETH exposure has cooled alongside the broader market.
• The AUM distribution continues to mirror the pattern established in Bitcoin ETFs, with BlackRock's ETHA holding the majority of assets throughout the drawdown. 
  • This suggests that even as aggregate AUM contracts, flows are consolidating toward the largest issuers, a dynamic worth watching as the market eventually recovers. In Bitcoin, the rise of issuers with lower fees may prompt some Ethereum issuers to consider the same.
• One potential differentiator for Ethereum going forward is its positioning relative to the emerging quantum computing threat narrative. Bitcoin's UTXO model carries meaningful exposure through old addresses with publicly visible keys, and no clear near-term upgrade path has been established. 
  • Ethereum, by contrast, has an active post-quantum roadmap, with Vitalik Buterin outlining migration paths and EIP-7560 addressing account abstraction in ways that could accommodate quantum-resistant signatures.
• Whether this translates into a meaningful capital allocation argument for ETH ETF investors remains to be seen. Institutional buyers in these products have historically been more responsive to price momentum and macro conditions than to protocol-level technical differentiation. But as the quantum narrative matures from theoretical to a more tangible concern, Ethereum's relative preparedness could become a more visible part of the investment thesis.

Competitive Markets 

• The Ethereum DEX aggregator landscape has shifted market structure over the past year, moving away from the near-dominance of a single protocol toward a more distributed competitive field. 
  • Kyber currently leads with approximately 31% market share, followed by CowSwap at 22%, while 1inch has seen its share decline from roughly 30% to 15% over the same period.
• This diversification stands in contrast to broader trends in crypto, where many sectors have seen consolidation. The all-time view shows 1inch commanding the majority of aggregator volume through 2021-2022, a period that coincided with aggressive liquidity mining campaigns across DeFi. 
  • The subsequent erosion of its share as incentives wound down raises a structural question of how much of any aggregator's volume reflects genuine user preference versus incentive-driven activity.
• It is worth noting that this data captures direct-to-aggregator volume only. If a user routes through aggregator A and it internally routes to aggregator B, only aggregator A receives credit. Flash loan trades are also filtered out. This methodology means the figures likely understate the reach of protocols that function as backends for other aggregators.
• Whether the current distribution reflects lasting structural change or another rotation driven by incentive cycles will become clearer as token programs mature. For now, the data suggests the aggregator space on Ethereum is more competitive than it has been in years.

Drift Trade Hacks

• On April 1st, Drift Protocol suffered a $285M exploit, placing it among the largest DeFi hacks on record, making it the largest non-bridge hack on Solana and a top-10 exploit by total funds stolen. 
  • Combined with prior exploits tracked on The Block's dashboards, the cumulative total stolen from DeFi protocols has now crossed $6 billion.
• What distinguishes this attack from typical DeFi exploits is not a smart contract vulnerability, but a structured, six-month social engineering operation. 
  • Threat actors successfully posed as a quantitative trading firm, built a genuine operational presence within the Drift ecosystem, and deposited over $1M of their own capital before executing the attack.
• Preliminary investigation attributes the attack with medium-high confidence to UNC4736, a North Korean state-affiliated group. The same group was responsible for the $1.5B Bybit exploit earlier this year, with overlapping on-chain fund flows and operational patterns tying the two attacks together.
• Taken together, the Drift and Bybit hacks illustrate a meaningful shift in how sophisticated attackers approach crypto infrastructure. The attack surface now extends well beyond smart contract code to the personnel, devices, and third-party relationships surrounding a protocol. 
  • As on-chain security matures, operational security at the human layer is becoming an equally critical and arguably underprepared line of defense.